Scholarlink

<p>Facial action unit (AU) recognition involves predicting the activation states of AUs, which describe facial movements. In complex scenarios, AU relationships and facial features are challenging to capture effectively. This study proposes a novel AU recognition approach that supplements topological features with AU relationship learning. By integrating a channel-topology convolution feature generation structure (CCFG) with a multi-scale attention feature generation structure (MAFG) within a graph neural network, our method models dynamic AU associations and enriches feature representations. Experimental results demonstrate that our approach significantly outperforms state-of-the-art methods on benchmark datasets, achieving average F1 scores of 66.6% and 67.2% on DISFA and EmotioNet, respectively, highlighting its robustness and precision in facial expression recognition tasks. Our code and dataset documentation are available at <a href="https://github.com/lkq52110/au-recognition">https://github.com/lkq52110/au-recognition</a>.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhancing facial action unit recognition through topological feature integration and relational learning

  • 期刊论文

  • 发表日期：

    2025年1月31日

查看全文
  • Keqi Li,
  • Yaping Wan,
  • Gang Zou,
  • Wangxiu Li,
  • Jian Yang,
  • Changyi Xie

摘要

Facial action unit (AU) recognition involves predicting the activation states of AUs, which describe facial movements. In complex scenarios, AU relationships and facial features are challenging to capture effectively. This study proposes a novel AU recognition approach that supplements topological features with AU relationship learning. By integrating a channel-topology convolution feature generation structure (CCFG) with a multi-scale attention feature generation structure (MAFG) within a graph neural network, our method models dynamic AU associations and enriches feature representations. Experimental results demonstrate that our approach significantly outperforms state-of-the-art methods on benchmark datasets, achieving average F1 scores of 66.6% and 67.2% on DISFA and EmotioNet, respectively, highlighting its robustness and precision in facial expression recognition tasks. Our code and dataset documentation are available at https://github.com/lkq52110/au-recognition.